tactical network planning for food aid distribution in kenya
TRANSCRIPT
Tactical Network Planning for
Food Aid Distribution in Kenya M.-È. Rancourt, J.-F. Cordeau, G. Laporte and B. Watkins, Computers &
Operations Research, 56: 68-83, 2015
Transportation Center Seminar Northwestern’s McCormick School of Engineering and Applied Science March 5, 2015
Marie-Ève Rancourt, Ph.D. Department of Management and Technology, ESG UQÀM
Outline
Context: humanitarian logistics
The network design problem
Field work and data collection
Mathematical formulation
Results
Conclusions and future research directions
Humanitarian logistics
The process of planning, implementing and
controlling the efficient, cost-effective flow and
storage of goods and materials, as well as related
information, from the point of origin to the point of consumption for the purpose of meeting the end
beneficiaries’ requirements
A. Thomas and M. Mizushima (2011)
Humanitarian logistics Disaster response versus development projects
Disaster
response
The Federal Emergency
Management Agency
(FEMA) defines a
disaster as:
« an event that causes
100 deaths or 100
human injuries or
damage worth 1 million
dollars »
Development
projects
Also involve human
suffering and
economic damage,
but covering longer
time-spans
Their cause can usually
not be traced back to
a specific catastrophic
event
East Africa struggles with…
Extreme poverty and rapid population growth
Wars and population migrations
Diseases (malaria, HIV/AIDS, …)
Gender issues and lack of education
Governance challenges
Fragile food production systems
Recurrent droughts and floods
Food insecurity
Food insecurity
Hunger and malnutrition are the greatest risks to global health (World Food Programme, UN)
Eradicating extreme poverty and hunger is the first goal of the eight UN Millennium Development Goals
Sub-Saharan Africa is the only region in the world suffering from persistent chronic food insecurity
Source: FEWS NET
Acute food
insecurity as of
today
Food aid as an instrument to reduce food
insecurity
Kenya
Between 1988 and 2011
182,000 MT per year on average (FAO)
Number of beneficiaries
14.3 million people in 2013-2014
Main causes
Poverty
Seasonal droughts
Refugee camps (about 480,000 refugees in Dadaab and Kakuma)
Food aid
Providing food and related assistance to tackle hunger,
either in emergency
situations, or to help with
deeper, longer term hunger alleviation and achieve food
security
This project focuses on in-
kind food donations to beneficiaries
Objective of this project
Objective: Improve the design of the food aid
distribution network taking into account the welfare of
multiple stakeholders
Scope: Determination of final delivery points, last-mile of
food aid distribution
Methodology: Mathematical programming
Problem class: Facility location and coverage problems
Geographical coverage: Garissa district, Kenya
Collaboration
The World Food Programme (WFP) of the United
Nations
The largest humanitarian agency, aims to fight against
hunger in the world
Know-how in the areas of food security analyses, nutrition,
food procurement and logistics (transportation and
warehousing)
Kenya Red Cross
Run different projects (services): famine, education, blood,
first aid, disaster and emergency
Scientific contributions
The main challenge of the project lies more in
modeling the problem, carrying out data collection
and processing, and performing analyses than on
algorithmic development
Describe the logistics processes of food aid distribution
and estimate stakeholders’ costs
First paper to apply optimization tools using real data in
the context of last-mile food aid distribution in Africa and
computing stakeholders’ tradeoff costs
Steps
1. Understand the food distribution process
2. Determine the network parameters Demand
Potential FDP locations
Distances
3. Estimate the stakeholder cost functions Beneficiaries
World Food Programme (WFP)
Kenya Red Cross
4. Formulate and solve the mathematical models
5. Estimate tradeoffs
Step 1: Understanding the food distribution process
Field work
Interviews
Facility visits
Food distribution observation
Food distribution process
Hub
Legend:
EDP
FDP
Food entries
(international
transport)
Extended Delivery Point
(EDP)
Final Delivery Point (FDP)
Secondary
transport
Primary
transport
Food aid regional supply chain Operations and stakeholders
Secondary
transport FDPs
WFP & Red-Cross Red-Cross &
Community
This project!
Beneficiaries
Operations
Stakeholders
Food aid Hand-out
(distribution) EDP Garissa
Garissa and its surroundings
Why Garissa and its surroundings?
One of the most vulnerable regions in Kenya
35% of the region’s population received food aid in
the last 12 years (62% during the most difficult period)
High poverty rate
Arid land with low rainfall
Pastoralism is the dominant livelihood system
Food aid is constant
Fixed distribution system which justifies the need for an
optimized network
EDP:
Red Cross is responsible for the
reception and storage of food
Red Cross organizes
secondary transport
operations, but the WFP bears
the transportation costs
FDP et la community
Facilitators:
“Food relief committee” chairman
Red Cross monitor
Spread settlements
Poor infrastructures!
18
Activities/Responsibilities at the EDP and a FDPs
“Community Relief Committee”
Elected by the community
Trained by Red Cross
Targeting, record keeping, arrange food distribution, provide storage and ensure security
Red Cross
Ensure that food assistance reaches beneficiaries
Assist the community
Activities/Responsibilities at a FDP
Food aid:
Vegetable oil
Sorghum (cereal)
Unloading
Truck arrival
Records: Beneficiary book
Distribution book
Activities/Responsibilities at a FDP
Activities/Responsibilities at a FDP
21
Shipment management Counting
Signing waybill
Losses/damaged bags
Activities at a FDP
Distribution “Scooping”
Hand-out (distribution)
Donkey transportation
service
Tactical “FDP” location problem
F
D
P
F
D
P
F
D
P
F
D
P
F
D
P
FD
P
F
D
P
F
D
P
Garissa
EDP F
D
P
Population points (V1)
Potential FDP locations (V2)
Transportation costs
(WFP)
Location and hand-out costs
(Kenya Red Cross)
Access costs (beneficiary
opportunity costs)
Costs:
Nodes:
Step 2: Determine the physical network structure
1. Demand
Population needs
Population locations
2. Potential FDP locations
3. Transportation network (distances)
Distance from each population point to closest road
Distance from Garissa EDP to each potential FDP
locations
Distance from each population points to each
potential FDP locations
Question 1 – Demand
Where are the beneficiaries?
Geographic Information Systems (GIS) and gridded
population data
How much food are they entitled to?
2012 Short Rain Need Assessment
Long rains Short rains Long dry Long
rains
Short dry
Need assessment: Determination of the demand for
the following 6 months.
Need assessment in Kenya
Need assessment in Kenya
For each division of Kenya, two parameters are
determined (effective for a period of 6 months):
Number of beneficiaries
Ration entitlement
Kenya
Locations (2,427)
Sub-locations
(6,612)
(# beneficiaries, ration entitlement)
Garissa
Districts (46 -- 16 to 26 )
Provinces (8)
Divisions (497)
400g of cereal flour/rice/bulgur
60g of pulses
25 g of oil (vit. A fortified)
50 g of fortified blended foods (Corn Soya
Blend)
15g of sugar
15g of iodized salt
Food basket
2012 Short Rain Assessment for Garissa and
its surroundings
Food aid requirement (tonnes/month)
Set of population points − V1 Source: GIS gridded population data
1 km
0 0 3 3 3
3 3
3
5 5 5 5
0 0 3 3 3 3
0 0 0 0 2 2
3 3 5 5 10
1
2
5 5
10
1
2
10
1
2
10
1
2 36
4
3
65
7
1 95
8
9
65
7
1
1
6
3
5
1 km
1 km
1 km
Needs qi at population point i :
With pi the population at i, B the number of beneficiaries,
P the total population and ration the entitled amount of
food aid per beneficiary at smallest division level
V1
qi =pi
PB × ration
Question 2 – Potential FDP locations
Where are the potential FDP locations?
Geographic Information Systems (GIS)
Road network
Population data
Set of potential FDP locations – V2 Sources: GIS gridded population data and road vectors
1 km
0 0 3 3 3
3 3
3
5 5 5 5
0 0 3 3 3 3
0 0 0 0 2 2
3 3 5 5 10
1
2
5 5
10
1
2
1
0
1
2
10
1
2 36
4
3
65
7
1 95
8
9
6
5
7
1
1
6
3
5
1 km
1 km
1 km
FDP • Close to a road (≤ 200 m)
• Population center (≥ 20 people)
V2 F
D
P
FDP
FDP
FDP
FDP
FDP
FDP
FDP
FDP
FDP
Question 3 – Transportation distances
What are the network transportation distances?
Geographic Information Systems (GIS)
Road network
Population data
Algorithms
Distances within the network
Garissa EDP to each
potential FDP
Road distances
Source: Google maps API
1460 distances
Each population point to each potential FDP
Geographical distances
Source: GIS
35,701,380 distances
3 3 3
3 3
3
5 5 5 5
3 3 3 3
2 2
3 3 5 5 10
1
2
5 5
10
1
2
10
1
2
10
1
2 36
4
3
65
7
1 95
8
9
65
7
1
1
6
3
5
FDP
FDP
FDP
FDP
FDP
FDP FDP
FDP FDP
Garissa
EDP
…
Network description
Step 3: Estimate the stakeholder costs
Stakeholders that bear costs
Beneficiary opportunity costs (access costs)
WFP (transportation costs)
Kenya Red Cross (location and hand-out costs)
Data sources
Beneficiary questionnaires
Contracts between the WFP and the Kenya Red
Cross
Beneficiary opportunity costs
Value of walking time:
0.25 h/km · 2 · distance to FDP (km) · 22,25 KSh/h
Value of food transport service (donkey):
20 KSh + 2.5 KSh/km · distance to FDP (km)
Beneficiary opportunity costs:
11,4 KSh/km · distance to FDP (dij) + 20 KSh
Minimum wage rate for unskilled labor Walking time (pace: 4 km/h)
Statistics based on a monitoring report for WFP
Transportation costs (WFP)
The Red Cross contracts and coordinates with local
transporters, but WFP fixes secondary transportation rates
and pays for the services:
Transportation costs to serve the FDPs depend on the
distance and the quantity of food delivered
F
D
P
Garissa
EDP
F
D
P
F
D
P
0 Q1
Q1
2 Q2
3 Q3
:
: ,
Q2
Q3
Location and hand-out costs
(Kenya Red Cross)
Fixed costs: Relief comity training and registration
validation
Two workdays for the Red Cross facilitator
Variable costs: Monthly food distribution
monitoring
Two workdays per month for the Red Cross staff
(announcement, dispatch and distribution)
Total estimated costs: KSh
Step 4: Mathematical formulation of the problem
Define the decision variables
Determine the objective function
Formulate the constraints
Decision variables and coverage
radius
Decision variables
Radius of coverage r and Wi(r)
F
D
P
F
D
P
F
D
P
Wi(r)
F
D
P
dij < r
d
ij > r
dij < r
dij < r
xij
xij
xij
V1(r)
≠ Ø
Mathematical formulation –
Cost Model
Beneficiaries opportunity
costs
Transportation costs
(WFP)
Location and hand-out costs
(Kenya Red Cross)
Open FDPs
Demand
Binary
Non negativity
Step 5: Computational results
Solve the problem using the CPLEX 12.5 library
in a C++ program Optimality gap: 0.1%
Comparative analyses
Impact of the response system structure on the
stakeholder welfare costs
Compare results of the cost model with classic covering models
Solution illustrations
Solution characteristics
% of the population covered as a function of r
Average walking time per beneficiary as a function of r
Covered people as a function
of the coverage radius
Uncovered people as a
function of the coverage
radius
% of the population uncovered as a function of r
Average walking time per beneficiary as a function of r
Assuming they register to the
closest open FDP
Stakeholder costs
Red Cross
5% of the total
cost on
average
Beneficiaries
22% of the total
cost on average
WFP
73% of the total
cost on average
Total welfare
cost
Stakeholder costs per beneficiary
Fair radius
Fair and cost-
efficient solutions obtained with:
r = 10, 11, … , 17.
Fair?
Complying with The
Sphere Project
Standards (2014), i.e.
90% of the
beneficiaries should
be covered within a
one-day return walk .
Here, about 92% of
the people are
covered with an
average walking time
of 2 hours.
Tradeoff between beneficiary and
transportation costs
Minimizing beneficiary
opportunity costs
Average % of decrease in
average walking time per
beneficiary
37%
Average % of increase in
transportation costs
14%
Minimizing supply transportation costs (WFP)
Average % of decrease in
transportation costs
15%
Average % of increase in
beneficiary average walking time
188%
A small increase in WFP costs can yield a large
reduction in beneficiary opportunity costs
Coverage Model
Maximize covered need with 156 FDPs
Comparative analysis – Coverage
Comparison of the % of covered people obtained with the cost model and the coverage model with 156 FDPs
Less covered people when r ≤ 10
Comparative analysis – Stakeholder
costs
Larger beneficiary and WFP costs for all r, but similar
cost when r = 10 km
Comparison of the stakeholder costs obtained with the cost model and the coverage model with 156 FDPs
Conclusions
Defined a framework to optimize food aid distribution networks (FDP locations)
Highlighted the importance of valuing the beneficiaries’ time
Found transportation costs to be the largest costs
Found that, taking beneficiary opportunity costs into account, a relatively low value of r minimizes total costs
Next steps:
How to design food aid supply chains that will lead to a more sustainable response and favour long-term economic growth?
Emerging aid systems Sustainable food security and resilient supply chains
« Cash and Vouchers »
Cash transfers provide money to people who are struggling to
provide food to their families
Vouchers can be redeemed for food items or « spent » in selected
shops
« Local purchase »
WFP purchases locally in developing countries in its criteria of price,
quality and quantity can be met
« Purchase for Progress »
Test new procurement approaches best suited for small producers
Support farmers to get better yields, reduce
losses, improve the quality of their crops and
connect them to markets
Future research
Dynamic and stochastic problem at the national level
Procurement: International
Local
Two type of commodities
Food
Cash & vouchers
Effect on local markets and food production
Stakeholders WFP and Kenya Red Cross
Beneficiaries
Local producers and traders
Non beneficiaries
Discussion
Questions and discussion…